Mickey Kaus has jumped on the anti-gerrymandering bandwagon. He quotes Will Vehrs, the Quasipundit, who quotes A. Barton Hinkle of the Richmond Times-Dispatch, and says "That makes about three of us!". Actually, Megan McArdle, Steven Hill and Rob Richie, and John Fund have already been on the case for almost a week. There may be others. As Hill and Richie say:
It will come as no surprise that, just like computers have impacted so many other areas of modern life, new computer technologies have dramatically altered the redistricting game.
Politicians and their consultants now have at their disposal extremely sophisticated computer technology, combined with the latest Census, demographic and polling data, to precisely gerrymander their districts. The days of plastic Mylar maps, Elmer's glue, magic markers, trial and error jigsaws and cut and paste blueprints are over. The software is more accurate than ever before, and the politicians have greatly enhanced capacity to handpick their voters.
Here's my question: If we already have all the data on disc, accurate down to the individual city block, why can't computers be programmed to draw the districts themselves? It seems to me that it would only be necessary to determine the criteria in advance, and then let the machines do the rest.
Criteria would need to be well defined, and ranked in ways that the programmers and their advisory committees found sensible. These committees would have to include statisticians, geographers, demographers, political scientists, perhaps some ordinary citizens, but absolutely no incumbents, staff members of incumbents, family members of incumbents, or pollsters.
Determining and ranking the criteria would be the hard part. Typical questions to answer would be:
There would obviously be a lot of questions like these. Examples could be multiplied, but that would be tedious, and is better left to the geographers and statisticians.
The Supreme Court has already decreed that it is more important to have equal districts than to have the boundaries follow (e.g.) county lines, so this is necessarily the primary criterion. The second and third should, I think, be contiguity and compactness, in that order. Compactness is easy enough to define unambiguously: which map has the shortest total length of district boundaries? This is a matter of more and less, one reason why it comes third rather than second. Contiguity is not much harder to define: can you get from one side of each district to the other without passing through any other districts? This is a pass-fail criterion. For example, any district which includes part or all of San Francisco and part or all of Oakland would also have to include both ends of the Bay Bridge and the inhabitants of Treasure Island and (if it has any) Yerba Buena Island, since they are in the middle of the bridge.
It would be possible to include shared ethnicity, occupation, or income levels among the criteria, so that the software would try to lump together (e.g.) Hispanics, or potato farmers, or the lower middle class. I see nothing wrong with that, as long as it is done without violating the three highest criteria. For example: A metropolitan area of around two million inhabitants must be divided into three equal districts. These could obviously look like three slices of pie, or they could look like one central doughnut hole with two half-doughnuts of suburbs around it. (To keep the numbers equal, the central district might have to include some inner suburbs, or the outer districts might have to include some outer bits of the city. But never both: that would be gerrymandering.) Given the geographic distribution of American ethnic groups, the doughnut arrangement would be more likely (all other things being equal) to increase the number of black or Hispanic congressmen. Perhaps it should therefore be preferred. Whether the suburbs of our hypothetical city should be divided east-west, or north-south, or diagonally, could also be determined by which choice would put like people together. Then again, if a major river flows through the area, it would make a more natural boundary. Again, I have no objection to putting the poorer, or whiter, or more rural suburbs together, as long as the districts are compact and contiguous, and as long as the criteria are set in advance. There is no need to go from today's ridiculously polarized districts to the opposite extreme and make them all such hodgepodges that only boring moderates could ever be elected.
One obvious objection to my proposal is that any software, and any set of criteria, would have to be tested to make sure they work, and possibly to tweak the ranking of criteria so that the results do not offend common sense and basic fair play. Since common sense and fair play are subjective concepts, there would be room for political prejudice to come in the back door, as it were. All the more reason to keep politicians off the advisory committees. It would also be important not to tell anyone on the committee where the incumbents live. Keeping incumbents out of each others' districts should be nowhere on the list of criteria. All in all, the committee might have to be sequestered while they do their work. Basically, the programmers would have to provide a list of possible criteria and examples of how they would work. The committee would rank and weight the criteria, the program would run once, and they would all go home, to reconvene ten years later. Actually, once the criteria were set, they would never have to reconvene.
In any case, I do not find this objection compelling. I doubt that it is possible to shuffle criteria of the kind I have outlined into any order or weighting that would produce anywhere near the results incumbents would want, or are used to getting. It takes the human touch to really 'fix' things.
There are also ways to combat the temptation to fiddle. One would be to allow (e.g.) Missouri redistricters to test their criteria on Maryland, and vice versa, but not on their own states. Better, the criteria could be tested using a different number of districts from the actual one. Massachusetts has 10 congressional districts. A program that could divide the state into 8, 9, 11, or 12 equal sections, each one compact and contiguous and generally reasonable looking, would presumably do the same when the target number was reset for 10 and the criteria were no longer adjustable. (I like to think of the politicians holding their breath as the software chugs along, waiting to see what district they will end up in, rather like participants in a professional sports draft. Make the bastards sweat: that's my motto.) Other safeguards could be proposed, for instance no criteria changes without a two-thirds or three-fourths vote of the advisory committee. I'm open to suggestion.
Of course, the main problem with my proposal is that it would be vehemently opposed by incumbents and their political consultants. The people who actually gather the data would still have plenty of work to do, and there would be added work for impartial advisory committees. But the consultants who spend their time plotting and scheming where to put the new district lines for maximum effect, helping their friends and harming their enemies all the way, would be out of a job. So would all the politicians unable to adjust to life without a guaranteed safe seat.
All in all, legislatures seem unlikely to take up my proposal, even if it can be proven to work. Perhaps a judge could order something like this to be done. Then again, perhaps not: a lot of judges seem to like the feeling of power they get from approving and rejecting individual redistricting plans in their states. So we can't count on them. And I'm none too comfortable with kicking legislative problems over to judges, anyway. Perhaps a referendum would be the way to go, at least in states that allow them.
In conclusion, can anyone tell me why this would not work? I don't claim my scheme is perfect, and it would certainly be politically difficult to implement, but it seems infinitely better than the current method.
Posted by Dr. Weevil at March 19, 2002 10:40 PM